In this discussion, Ram Neta says (sort of) that the people who have argued most recently contingent a priori knowledge aren't internalists. I pointed to a couple of people who seemed to me to be internalists who might accept contingent a priori knowledge. But perhaps what I should have said was this:
I'm pretty much an internalist; and the part of me that cares about knowledge, and that cares at all about the a priori, believes in contingent a priori knowledge.
Do you think there's something funny about the thing I think I should have said?
Well, for an epistemologist to say that only part of him cares about knowledge may be funny. (Or maybe not--I've been told that the majority view is that knowledge isn't that important.) But that's not what I was getting at. The sentence seems grammatically OK to me, but it contains a negative polarity item, "at all," within a definite description, "the part of me that...." There isn't any other obvious NPI-licenser in the vicinity. And definite descriptions usually don't license NPIs.
(That Daniel Rothschild paper via Weatherson. The longer version seems to have been taken down, though I could swear I printed it out the other day. For a quick account of what NPIs are go here, though some of the claims are disputed; particuarlly the one at the end about downward entailment.)
I'm sort of tempted by this response: when I talk about "the part of me that cares at all about the a priori," I'm implicating that I don't really care about the a priori. So the NPI is licensed. That's free pragmatic enrichment of a kind that I understand to be unpopular among people who think about it; and though I'd perversely kind of like it to exist, I'm not going to argue about it until I think about it some more.
Here's another, maybe more promising response. Rothschild points out that NPIs are licensed in descriptions that make the uniqueness condition explicit, as in "The only person who understood me at all didn't understand me." And you could argue that "the part of X that Y" actually does, in virtue of its semantics. The part of me that cares about the a priori equals all and only those constituents of me that care about the a priori. (Or something like that. The part of me that knows about mereology is even smaller.) By definition there's only one thing that can be denoted by "the part of me that cares about the a priori."
(Though I'm not sure about this--it probably already incorporates tendentious assumptions about 'the'. I can say "A part of me that cares about the a priori believes in contingent a priori knowledge, and another part of me thinks it's nuts." Um, less technical example: "After I fell down, there was a part of me that was covered in mud that was all scraped up, and another part of me that was covered in mud that wasn't.")
Consider this sentence:
The part of Texas that gets any snow at all doesn't get as much snow as Milwaukee.
That doesn't have the pragmatic implication that Texas doesn't really get any snow at all. Is the sentence acceptable? I kind of think so. And it might be taken to be neither downward- nor upward-entailing; the part of Texas that gets snow is different from the part of Texas that gets snow or rain (it's smaller) and from the part of Texas that gets snow and hail (it's bigger). So maybe some implicit uniqueness in "the part of" answers the question. I don't know. As you may see a couple of posts later, I am confused by NPIs.
(Thanks to Allan Hazlett for discussions of some of this--any mistakes are mine, as usual.)
Posted by Matt Weiner at September 7, 2005 06:06 PM"The part of Texas that gets snow" implies that there is a part of Texas that doesn't get snow.
"The part of Texas that gets any snow at all" implies that that part of Texas that doesn't get snow, doesn't get any snow at all... not even a little bit.
Likewise by bringing up the part of you that cares at all about the a priori, you are clearly saying that part of you does not care at all about the a priori. (1) I believe this implied negative statement licenses the NPI. It is not implied nor is it necessary to imply that you, the whole of you, does not care about the a priori.
(1) N.B. I have no idea what you mean by that, but it's clearly what you're saying.
Posted by: Richard Mason at September 8, 2005 09:25 AMI'd say those are implicatures rather than implications, FWIW. But I think the difference between the two 'part of Texas' locutions is that when I say "The part of Texas that gets snow" I may mean everywhere in Texas that gets more than a little bit of snow, and when I say "The part of Texas that gets any snow at all" I mean everywhere in Texas that gets any snow at all. Like, if Brownfield gets a dusting of snow every couple of years, it might be in the part of Texas that gets any snow at all without being in the part of Texas that gets snow.
But this may not be literally true; maybe literally, the part of Texas that gets snow just is the part of Texas that gets any snow at all, and when I don't include 'at all' I convey that the amount of snow involved is significant. I don't know--I'm definitely no expert on part-talk.
To the substance of your proposal--I think that's like the first one I consider, in which the NPI is licensed because I'm implicating a negative. But I'm worried that the proposal proves too much. If I say "the person in this room with any hair at all is drinking a martini," I'm at least implicating that most of the people in this room don't have any hair at all. According to your proposal (as I understand it), that should license the NPI. But, as Rothschild points out with this example, the NPI isn't licensed--"The person in this room with any hair at all" sounds funny.
(One think Allan pointed out that I didn't mention is that it's important to distinguish the case in which the 'the'-phrase is a quantifier--"The people in this room with any hair at all are all leaving by the east door" sounds better, but that's really a universal statement, in which we'd expect NPIs to be licensed.)
What I meant by the thing about the a priori is kind of like this: Officially and sometimes in practice, I think specific questions about knowledge aren't too important. Sometimes when I'm confronted with an example and asked "Does X know?" I find it hard to care about whether X knows. But I find it easy to think myself into a position where I care about those questions--so I say there's a part of me that cares about knowledge. I find it harder, but not impossible, to think myself into a position where I care about questions of whether this particular thing can be known a priori. So what I'm saying is that I don't really care about the a priori, and the part of me that cares about it doesn't care about it very much.
Posted by: Matt Weiner at September 8, 2005 11:23 AMI think that's like the first one I consider
Yes, it is. But I was reacting to your examples, in which you seemed concerned about whether you cared about the a priori, and whether Texas really gets any snow. The negative implication doesn't have to be about you or about Texas taken as a whole.
I still think this was a good nitpick, but perhaps no more than that, i.e., perhaps the first hypothesis still isn't right. I'm not sure now.
"The person in this room with any hair at all" sounds funny.
I don't think it sounds all that funny.
(a) The person in this room with any hair at all may be in the habit of carrying a comb, but the rest of us aren't.
(b) The person in this room with any money at all is the least likely to want to contribute.
(c) The person in this room with any hope at all of fitting through the ventilation shaft has just collapsed in a drunken stupor.
(d) The person in this room with any trace at all of gunpowder residue on his hands must be the murderer.
These, especially (b) and (c), probably would sound better if a word such as "one," "only," or "single," were explicitly used, but I don't think that's absolutely required. I'm not even certain that (a) or (d) would be improved.
As you note, if we pluralize the subject then all these sound fine.
Posted by: Richard Mason at September 8, 2005 04:10 PMMm, the second one wasn't actually supposed to imply that Texas doesn't really get snow. The difference between it and the first is that in the first one can say, "So you don't really care about the a priori," and in the second you can't say "So Texas doesn't really get snow"--if a part of Texas gets snow, then Texas does. I didn't make that clear. (What happened was, I thought "The part of me that cares...", realized there was something funny about the NPI, and then came up with the Texas example when I was thinking about whether the NPI-licensing was due to the negative implicature.)
In your examples (b) and (d) strike me as quantificational. At least, when you utter them, you're not likely to base them on specific information about that person, as opposed to (a) and especially (c). Don't know if that makes a difference. At the moment I'm kind of wonky about whether I think (a) is OK--thinking about these for long enoughs makes my intuitions go all funny.
Posted by: Matt Weiner at September 8, 2005 04:34 PM